Karin Cox
Learning Research and Development CenterRoom 613
3939 O'Hara St.
Pittsburgh, PA 15260
TEL: 412-624-3121
FAX: 412-624-9149
EMAIL: kmc51@pitt.edu
I have been a graduate student in the Fiez lab since 2004. My research aims to address the question of how humans learn from the positive and negative consequences of their actions. Such outcome information can be applied to very simple tasks, like tracking the rate of reward associated with different options. Outcome information also allows people to learn about more complex relationships, and my most recent work focuses on this kind of learning.
I am currently interested in a phenomenon known as the "discrimination learning set," in which humans (and other species) use outcome information to learn about the structure of a particular task domain. In a deterministic stimulus discrimination problem, subjects must learn to choose the always-rewarded "S+" stimulus as opposed to a never-rewarded "S-" stimulus; after exposure to only a few unique problems, adult human subjects will learn to reliably approach or avoid a novel stimulus following the receipt of a single reward or nonreward outcome. I have recently found that humans can also use a single reward or nonreward to infer the value of a stimulus that was not chosen on the first trial of a novel problem (e.g., if choice of stimulus A was followed by reward, then stimulus B must be associated with nonreward). This result is interesting in light of the fact that most models of reinforcement learning cannot account for value predictions that are based on inference, as opposed to direct experience. Therefore, I am using both behavioral and neuroimaging methods to explore the potential psychological and neural mechanisms underlying such value-related inferences.